Methodology For The Comprehensive Study Of A Multidimensional Statistical Sample In The Diagnosis Of Spinal Diseases

Keywords: lumbar spine, diagnosis of osteochondrosis, multidimensional data, cluster analysis, neural metric, sensitivity analysis, decision formalization

Abstract:

The report proposes a methodology for the analysis of statistical samples of multidimensional data. According to this methodology, the centers of the clusters in the studied statistical sample are determined through comprehensive application of cluster analysis methods. These cluster centers are associated with values of neurogrowth metrics, Euclidean metric, and a combined metric defined as the product of the neurogrowth metric and the Euclidean metric. The methodology is demonstrated on a statistical sample used in the diagnosis of spinal diseases. The task is to identify which factors (intervertebral distances in the lumbar spine: z₁, z₂, z₃, z₄) most influence the transition of patients and their diagnoses from one cluster to another. Sensitivity analysis of the metrics to variations in the intervertebral distances z₁, z₂, z₃, z₄ was performed, showing a high sensitivity to changes in z₂ (probability 100%), substantial sensitivity to z₁ (88%), and low sensitivity to z₃ (60%). This methodology enables the identification of factors responsible for transitions between patient clusters, thereby contributing to improved early diagnosis methods for osteochondrosis.
Keywords: lumbar spine, diagnosis of osteochondrosis, multidimensional data, cluster analysis, neural metric, sensitivity analysis, decision formalization.

Submission Number: 138